International growth
Several markets at once is a different problem from entering one
Past a couple of countries, the failures stop being about demand and start being about interference: English versions competing with each other, annotations that contradict themselves, prices that do not match the visitor, and reporting that averages every market into one unreadable line.
Recognise the stage
The symptoms of an estate that has outgrown its structure
These arrive somewhere around the third market and get worse with each one added. If four or more of them are true, the constraint on growth is your operating structure rather than demand in any individual country.
- A page built for one country regularly appears in the results of another
- Several near-identical English versions exist and you cannot predict which one surfaces
- Adding a page in one market has, at some point, made another market disappear
- Prices, delivery promises or payment options shown do not match the visitor’s location
- Visitors are automatically redirected to a version they did not choose and cannot easily leave
- Each country manager reports a number head office cannot reproduce
- Local teams have bought their own analytics, call tracking or automation tools
- Nobody can say which market a piece of group content was written for
- The same product has three different names across your own sites
Four interference failures
What each failure looks like and what actually resolves it
These are the specific mechanical failures behind most international underperformance. All four are fixable, and none of them is a strategy problem.
| Dimension | What you see | What is actually happening | What resolves it |
|---|---|---|---|
| English variants competing | The wrong country’s page ranks, or one version dominates everywhere | The versions are too similar to distinguish, so one is chosen and the others suppressed | Complete reciprocal annotations, plus genuinely local content in each variant |
| Annotations silently broken | A market that used to appear stops appearing after a routine publication | A page was added without return references, removing it from the set | Automated validation on publish, and a periodic audit of the whole estate |
| Currency and fulfilment mismatch | High traffic and very low conversion in one or more markets | The visitor is shown a price, delivery promise or payment method that does not apply | Market detected from the URL structure, with a visible and changeable choice |
| One averaged report | Group performance looks flat while individual markets swing wildly | Every market is blended into one line with one definition and one currency | Per-market figures against a group benchmark, with demand and conversion split |
Underneath the mechanics
Three structural drifts that keep recreating the mechanics
Fixing the four interference failures is a project. Keeping them fixed requires addressing why they appeared, and it is almost always one of these three.
- Markets were added one at a time by different people.
- The first market was a considered decision. The fourth was a folder someone created on a Thursday. Each addition inherited conventions from whoever was available, so the estate now contains three different URL patterns, two ways of handling language, and a set of annotations nobody has understood since the person who built them left. Nothing is wrong enough to force a rebuild and everything is wrong enough to cost visibility.
- Nobody owns the estate, only its parts.
- Each country manager owns their market and is measured on it. No one is accountable for the interactions between markets, which is exactly where the failures live. A page published in one country that damages another is nobody’s incident, so it is never investigated, and the same class of problem recurs indefinitely.
- Group content is written for no particular market.
- To avoid favouring one country, group content is written in an accent-free register aimed at everybody. It lands as foreign everywhere, answers no market’s specific objections, and is the material most likely to duplicate across variants. Content that is nobody’s reads as nobody’s, and it is the least productive output in most international estates.
The governance question
Decide ownership function by function, not as a philosophy
Almost every international organisation argues about centralisation as though it were a single question with a single answer. It is not. It is a series of separate decisions, and the right answer differs for each because the cost of being wrong differs for each.
Some things have to sit with the group, and the reason is exposure rather than control. What the company claims about its product, how regulated wording is handled, what a conversion is defined as, and which analytics configuration produces the group view — inconsistency in any of these is either a legal problem or a reporting problem, and both of them arrive at head office regardless of where the decision was made.
Other things have to sit locally, and the reason is that head office is reliably wrong about them. Which platforms matter in a given country, when the season falls, what tone reads as credible rather than presumptuous, which partnerships carry weight, what a fair price looks like locally. These are not preferences to be tolerated; they are the operating knowledge that makes the market work, and overriding them from a different country is how a group loses a market it had.
The failure mode is not choosing either extreme. It is never writing the division down, so that every disagreement becomes a negotiation about authority rather than a question about a specific decision. A single page listing which functions are group-owned, which are market-owned, and who resolves a conflict, removes more friction than any restructure.
One caveat worth stating plainly. A group standard that provides nothing gets ignored, and enforcement across borders is slow, political and rarely successful. Standards that come attached to something the market wants — a working analytics setup, a content budget, creative they would otherwise have to commission — are the ones that get followed.
The order of repair
Fixing a multi-market estate without stopping it
This sequence assumes you are trading throughout. Nothing here requires a rebuild, and the first two stages usually account for most of the recovered performance.
Establish what actually exists
Every market, every language version, every URL pattern, every annotation, every analytics property and every tool a local team has bought. This is rarely documented anywhere and the reality is always more complicated than the diagram. Expect to find two markets nobody at group level was tracking.
You get: A verified inventory of the whole estate
Validate the annotations across every version
Check that every version references every other and that each reference is returned. Fix the breaks, then automate the validation so that publishing a page cannot silently remove a market from the set. This is the highest-value technical work available in most estates and it is unglamorous enough to keep being deferred.
You get: A complete, reciprocal annotation set with validation on publish
Make each variant genuinely local
For markets sharing a language, the variants have to differ in substance rather than in currency symbol: local proof, local objections, local terminology, local contact routes. Where a market cannot justify that effort, consolidating it into a regional version is a better outcome than maintaining a near-duplicate.
You get: Variants with real local substance, or a deliberate consolidation
Align the commercial layer with the visitor
Currency, tax display, delivery promises, payment methods and stock availability matched to the market the visitor is actually in, determined by the section of the site rather than only by their browser. Keep the choice visible and reversible, and stop redirecting people automatically.
You get: A commercial layer consistent with each market section
Agree one measurement definition, then rebuild the reporting
One definition of a conversion, one analytics configuration, one group currency with local currency alongside. Then per-market figures against a group benchmark, with demand and conversion separated. Most reconciliation disputes are about definitions, so the agreement has to come before the dashboard.
You get: A group reporting model both head office and markets accept
Write down who owns what
One page: group-owned functions, market-owned functions, and the named person who resolves a conflict between them. Attach something the markets want to the group-owned side, because standards that supply nothing are standards nobody follows across a border.
You get: A one-page ownership model with a conflict route
Where the work sits
The four workstreams behind a working estate
Structure and annotations
Local substance
Questions
What group marketing teams ask us about international estates
Why do our English pages compete with each other?
Because they are close to identical. A French and a German version of a page are obviously different documents; four English versions differ by a currency symbol, a phone number and a spelling. There is very little for a search engine to tell them apart by, so it picks one and largely suppresses the rest.
Language and region annotations exist precisely to resolve this, and they only work if they are complete and reciprocal. Where they are incomplete, the default behaviour returns and the wrong market gets the wrong page.
We implemented hreflang years ago. Why is it still wrong?
Because it is not a one-off implementation, it is a state that has to be maintained. Every version must reference every other version and each must reference back. Add one page to one market without its return references and that page silently drops out of the set.
Nothing in a content management system warns you about this, and it accumulates. Any estate that has been publishing for a couple of years across several markets should assume the annotations are partially broken and check rather than believe.
Should marketing be run centrally or by each country?
Both, decided function by function rather than as a principle. Brand claims, regulated wording, measurement definitions, the analytics stack and the reporting benchmarks belong to the group, because inconsistency there creates legal exposure and makes the numbers meaningless.
Channel selection, creative execution, local partnerships and promotional timing belong to the market, because the platforms and behaviour genuinely differ by country and head office reliably underestimates by how much. Write the division down; the arguments happen where it was never made explicit.
Do we need a separate site for each country?
Usually not. Most multi-market businesses are better served by one well-structured estate with clear region and language sections than by several independent sites, each of which needs its own authority, its own maintenance and its own budget.
Separate country domains earn their cost when a market has genuinely distinct operations, its own legal entity, its own catalogue, and enough scale to fund the visibility from scratch. That is a smaller number of businesses than the ones who have chosen it.
How should international performance be reported?
With every market figure shown against a group comparison for the same period, and with demand separated from conversion. Without the group line you cannot tell a weak market from a weak quarter, and that distinction is the entire point of group reporting.
One definition of a conversion, one analytics configuration and one currency for the group view, with local currency alongside it. Most reconciliation disputes are definitional rather than numerical, and they are resolved by agreeing terms rather than by rebuilding dashboards.
Our prices show in the wrong currency. Is that a real problem?
It is one of the most direct conversion losses available to fix. A visitor shown a foreign currency, a delivery promise that does not apply to them or a payment method they do not use will leave, and no amount of upstream visibility work compensates for it.
Detect the market from the URL structure rather than only from the browser, keep the choice visible and changeable, and never trap someone in a version they did not pick. Automatic redirection based on location is the most common way this gets made worse.
Have the whole estate checked, not one market
Send us your market list and site structure. We will validate the annotations across every version, find the pages competing with their own variants, and tell you which of your markets is underperforming because of interference rather than because of demand.
If we don't deliver the work we agreed to deliver for reasons within our control, you don't pay for the undelivered work. Read our guarantee
Last updated · Reviewed by Zubair Afzal